Connect Anything
Merge agent skills into one, with provenance and an eval gate
You already run two or three skills for one job. Each asks to be invoked, loads its own context and repeats rules the others also have. Combining them by hand loses the reasons, the licences and the conflicts. Connect Anything does the merge as a build with evidence, and writes the result only when it matches or beats running the originals.
npx skills add IndusCore-AI/connect-anythingWorks in Claude Code, Codex CLI, Cursor, GitHub Copilot, claude.ai. Needs Python 3.9 or later and git, no packages.
Ask the way you would ask a person
- connect apple-design and dashboard-design into one skill for review work
- compile one research skill from this skill-radar report
- my apple-design skill is weak on app screens, update it
What Connect Anything does differently
Every file is read in full
Side files a SKILL.md never mentions included, with a hash of each.
Every conflict is named
MERGE.md shows what one source says, what the other says, which one wins at this job and why.
Every line has a way home
connect-anything.lock.json records each source's commit, licence, read count, scan result and kept files.
A merge that loses is refused
The eval gate runs on every build. The compiled skill is written only when it matches or beats the originals.
The sources are never trusted
Read as data, scanned line by line, never executed. Download-and-run, secret reads, hidden text and prompt injection are excluded and named.
No model call without a budget
The static tier is free and runs every time. The live tier needs --max-cost-usd.
Three entrances, one engine
| Skill | Give it | Get |
|---|---|---|
| connect-skills | The two or three skills you run together for one job | One skill for that job; the originals stay installed |
| compile-skills | 5 to 20 candidates (a skill-radar report, an awesome list, a search) and a profile | One skill made of only the parts that fit |
| update-skills | A skill you already use, and a gap you name | Upstream changes re-read and re-scanned, candidates for the gap, the eval re-run, nothing written until you approve |
- 01
Fetch each source at a locked commit, read every file in full and scan every line after normalising away hidden characters, look-alike letters and two-step downloads.
- 02
Check the licence. MIT, BSD, ISC, Apache-2.0, CC-BY-4.0, CC0 and Unlicense can be copied; no licence, NonCommercial and NoDerivatives are named and never copied.
- 03
Merge one SKILL.md with a what-wins list, a folder per source, MERGE.md and a lockfile, then run the checks and the eval gate: emit, or refuse with the scores.
Real runs, committed in the repo
Install it in your agent
npx skills add IndusCore-AI/connect-anything/plugin marketplace add IndusCore-AI/connect-anythinggh skill install IndusCore-AI/connect-anything connect-skillsWhat it does not do
- Unlicensed sources are named and left for you to install, never copied.
- It does not install skills or do hash-only updates of installed packages; npx skills update does that.
- A live eval run needs a cost budget agreed first. A dry run plans the calls and costs nothing.
Questions
Three open-source Agent Skills on one engine: connect-skills merges the skills you run together, compile-skills builds one skill from many candidates, and update-skills refreshes a skill you use. Each keeps provenance and an eval gate.
npx skills update compares hashes of installed packages. Connect Anything reads and re-scans every changed file, keeps your local edits, can find skills that fill a gap you name, and re-runs the eval before proposing anything.
Sources are read as data and never executed. Every line is scanned for download-and-run steps, secret reads, hidden text, prompt injection and forced promotion; matches are excluded and named.
Nothing is written to the output folder. The scores and the reasons go to REFUSED.md in the work folder.
Python 3.9 or later and git. No packages, and no network at run time except fetching the sources. It is MIT licensed and built by IndusCore.
Also from IndusCore: Intelligent UI Anything
An open-source agent skill that turns what Claude Code, Codex or Cursor produced into one interactive page, and sends your pick back to the agent.